Molecular Recommender · ProofNoMore

What else should a ProofNoMore customer reach for?

Pick a drink. The engine breaks it down to its flavor compounds and ranks the rest of the range by how much molecular ground they share — the quantitative "you'll like this" you described, running on your own shelf.

How it scores
Each drink is reduced to its flavor-driving ingredients (off the public label). Similarity = the average best-match affinity between the two drinks' ingredients, where each ingredient-pair affinity is the share of flavor compounds they hold in common (Jaccard), straight from the engine. Recipe counts are real co-occurrences from CompKitchen's 432,799-recipe co-occurrence index. Ingredient-level only; the finished carbonated liquid would need GC-MS.

Powered by the CompKitchen engine · slice = 7 of ~550 SKUs · full catalog build under NDA.